Banking Financial Data MCP Server

This is a Model Context Protocol (MCP) server that provides AI agents with access to simulated banking and financial data. It allows AI models to retrieve account balances, query transaction histories, analyze spending, manage cards, and more.

Built specifically to integrate with Mindset AI using the streamable-http (JSON-RPC) transport.

Features

Provides 14 specialized financial tools for AI execution:

1. get_account_balance: Check balances for savings, current, salary, and wallet accounts.
2. get_transactions: Retrieve paginated transaction history, filtered by date, type, or category.
3. get_last_15_days_statement: Fetch a focused 15-day bank statement.
4. initiate_transfer: Transfer funds to accounts or beneficiaries (Requires Mock OTP: "123456").
5. pay_bill: Pay utility bills to registered billers (Requires Mock OTP: "123456").
6. get_spending_analysis: Get AI-categorized spending breakdowns and insights (food, shopping, etc.).
7. get_loan_details: View active loans, total outstanding principal, and upcoming EMIs.
8. get_investments: View investment portfolio including Fixed Deposits and Mutual Fund SIPs.
9. manage_card: Block, unblock, or set limits for debit and credit cards (Requires Mock OTP).
10. change_debit_card_pin: Securely change a card's PIN (Requires Mock OTP).
11. set_budget_alert: Configure monthly spending budgets and alert thresholds.
12. get_net_worth: Automatically aggregate all assets minus liabilities to calculate total net worth.
13. get_financial_health: Get a composite financial health score system.
14. ask_support_faq: Highly accurate fuzzy-search into the banking FAQ database for customer support queries.

Database Architecture

This MCP server natively supports a Dual-Database Layer:

1. In-Memory Mocking: High-speed, stateless sandbox data loaded dynamically into RAM. Perfect for immediate testing.
2. MongoDB Atlas Support: Production-grade async motor database routing. If MONGO_URI is provided, the server securely streams queries directly to your Atlas collections. Includes large-scale test generation (generate_scale_data.py).

Prerequisites

• Python 3.12
• Vercel CLI (optional, for deployment)

Local Development Setup

1. Clone and setup a virtual environment:

python -m venv venv
# On Windows
venv\Scripts\activate
# On macOS/Linux
source venv/bin/activate

1. Install dependencies:

pip install -r requirements.txt

1. Configure Environment Variables: Create a .env file from the example:

cp .env.example .env

Ensure you generate and set a secure MCP_API_KEY in the .env file.

1. Run the server locally:

# Uses uvicorn to serve the FastMCP ASGI app on port 5000
uvicorn server:app --host 0.0.0.0 --port 5000

Testing Mindset AI Handshake Locally

The repository includes a test_integration.py script that perfectly simulates the 4-step MCP handshake executed by Mindset AI's web platform.

1. Keep the local server from the previous step running on port 5000
2. In a new terminal, run:

python test_integration.py

You should see all responses return HTTP 200/202 indicating a successful integration.

Vercel Deployment

This project is fully configured for Vercel deployment as a serverless Python function.

1. Push your repository to GitHub.
2. In the Vercel Dashboard, import the repository.
3. Keep default settings (Framework Preset: "Other", Blank Install/Build Commands).
4. In Environment Variables, add:
   • MCP_API_KEY: <your_generated_secure_api_key>
5. Deploy!

Vercel will use the provided vercel.json and runtime.txt configs automatically. Your Live MCP Endpoint will be: https://<your-vercel-domain>/mcp

Mindset AI Registration Context

When registering this server inside Mindset AI, use the following:

• Server URL: https://<your-vercel-domain>/mcp (must securely resolve via HTTPS)
• API Key: Provide the secret key you configured as MCP_API_KEY
• Transport Type: HTTP (Stateless)

Mindset AI's Agent Sessions will automatically pass user contexts (x-user-id, x-session-tags) to the endpoint which the tools consume to scope user data appropriately.

Technical Stack

• FastMCP: Recommended high-level MCP Python framework.
• Starlette/Uvicorn: ASGI server for web handling and routing.
• [Motor/PyMongo]: Asynchronous MongoDB driver for massive scaling.
• Vercel Serverless: Cloud deployment backend.
• Python 3.12: Runtime target.